You buy it directly from nVidia for £1099 - just got my one through yesterday.

 

Interesting. There it is just entitled "Titan X" not "Titan XP" or "Titan X Pascal", but it says Pascal architecture in the smaller print. So are ALL titan X's with Pacal? nVidia don't appear to list the Titan X separately like resellers do. 

 

So, maybe I was searching for the wrong thing. Amazon has Titan X's as low as £899, eg: the one below.

 

[LATER] Read the description of that card in the more detailed text below on Amazon -- it says "Maxwell" architecture".. So now I'm really confused. When buying a Titan X how do you know what you are getting? Does it say Pascal clearly on the box? It doesn't in the pic on a posting above.].

 

 

 

MSI Nvidia TITAN X Graphics Card (1000 MHz, GDDR5, 384 Bit, DL-DVI-I)

by MSI

£899.00Prime

Pete,

 

What type of Titan X , Maxwell or Pascal is written on the box..

 

I have a Maxwell Version. Depending on the resolution you use ( 2k or 4K ) , the AA settings and other intern P3D settings related to the GPU, you will see an improvement.

Rob claims about 50% , but I think in my case perhaps 15-20% with my config.

 

I see no reason to upgrade from a Titan X Maxwell to a Titan X Pascal.  

But you decide for yourself.

 

Benchmarktests :

 

http://www.babeltechreviews.com/battle-titans-pascal-titan-x-vs-maxwell-titan-x/3/

Ran my standard benchmark test I've used over the years for P3D - see tests 26,34,35,36:

 

Hi Rob,

 

Interesting set of results. However, I couldn't help noticing that test #29 compares very favourably with your latest batch

 

(Note: Those with sharp eyes will notice that Rob tried my 'rogue' affinity mask value of 21760 with his 8-cored CPU without HT. Rob detected more long frames which hadn't been my experience, although the difference between our results were marginal. It remains unclear why my experience had been better as our Setups, at that time, were very similar.)

 

Regards,

Mike

What type of Titan X , Maxwell or Pascal is written on the box..

 

Okay. Hopefully always in the adverts too then, though not necessarily very prominently it seems!

 

Odd that on the UK nVidia site only the one Titan X is listed, and that's just called "nVidia Titan X" just like many of the ads. The one there is a Pascal. I assume they are no longer supplying the Maxwell version so the resellers are just supplying older stock.

 

Thanks for the clarification. I have only recently got my 1080,. I see no need to upgrade again yet -- maybe if and when I decide to upgrade my projector to 4k.

 

Pete

Some of the discrepancy in these benchmarks may have something to do with the Bottleneck of a Third Kind. The first type is when the CPU speed limits the GPU. The second is the reverse. But the third kind is when the app needs more VRAM than is physically present on the graphics card. On lower end graphics cards it could occur at lower settings.

 

It is distinctly possible when running at 4K resolution with 4096 uncompressed 32 bit textures that this third type of bottleneck occurs. At that point the GPU starts using slower system RAM in addition to its faster onboard VRAM.

 

XP, for example shows VRAM usage when one adjusts texture complexity. 3rd party monitoring apps like Aida 64 will also display used VRAM on the fly.

 

 

Does it say Pascal clearly on the box?

 

Yes, it will clearly say "Titan X Powered by Pascal" on the box.

 

I know your an AI traffic kinda guy, that's mostly CPU intensive ... a Titan X Pascal isn't going to help much there.  More CPU and higher clocks will help for AI traffic.  In my testing .cfg link I provided, you'll notice that I have AI turned off, this is done so that testing is not skewed one way or the other (CPU/GPU) and my attempt to retain a balanced "flight".  It's the same test I've repeated over time as to give a relative performance compare.

 

For example, if use Orbx SoCal with FSDT KLAX and turned up AI, I'm immediately CPU bound and it wouldn't make much of difference what GPU I used.

 

Cheers, Rob.

 

 

However, I couldn't help noticing that test #29 compares very favourably with your latest batch

 

Test 29 is with SLI enabled and DR is turned off relative to test 34,35,36 which are single GPU with DR at Medium.  DR is a good test of GPU performance, Test #26 is a more accurate compare to test 34,35,36.

 

Long frames will be very specific to the flight path and location and add-ons ... I'll record my test flight and post video in case you'd like to use it for reference.  Part of that test flight is to fly thru a cloud as we know that clouds and AA settings can stress a GPU.

 

Cheers, Rob.

Why do you need that 12 Gbs Rob?  Am I getting a whiff of 64bits?  What are you guys cooking up and savoring over there?   

Actually, just want to say Happy Holidays and thanks for doing what you do.  I tend to get a little bit of cheer when I see your savory posts.  Thanks and Cheery Holidays..

Hey Rob,

Been away from the forums for a few days (darn work stuff!) but thanks for your reply to my earlier post (towards the top) and the additional config and scenario data.  If I have some time this weekend I'm going to set up some similar test scenarios on my system and see what kind of results I get.

 

Btw I wasn't in any way implying that I didn't believe your results...I know how methodical you are with your test methods!  :wink:  I've had a nagging suspicion for some time that I've got a bottleneck in my system somewhere and that perhaps I'm not getting all the performance that might be available...not just in P3D but across the board.  Or maybe I just expect too much...anyway this might help confirm one way or the other or at least give some clues.

 

Happy Holidays!

Kevin

 

 

Btw I wasn't in any way implying that I didn't believe your results...

 

Not looking for believers, I just present my information ... if it's helpful great, if not, sorry, but that's ok too.

 

The problem with "bottlenecks" is that it will vary based on what needs to be rendered at any given time and location.  If one runs SGSS AA at high levels with dense clouds then the GPU is going to be the bottleneck but on a clear day SGSS AA works well.  If one runs high levels of AI and road traffic in a dense populated area like KSEA then CPU is going to be the bottleneck.

 

 

 

Why do you need that 12 Gbs Rob?

 

Well I had 24GB VRAM with my two (now "old") Titan X's.  But so far P3D/XP11/AF2 have not used more than 5.1GB VRAM ... but XP11 has used 16.1GB of VAS but ironically less VRAM?

 

Cheers, Rob.

 

Yes, silly. See Amazon entry below.

 

There may be other traders (probably doing "grey imports"  -- dodgy warranties that way) for less than that, but these things are always much more expensive over here! They may be trading on its relative rarity in the UK to date.

 

Anyway, I'm okay at present with my 1080, at least in FSX-SE. May consider upgrading later, maybe when I do eventually switch to P3D.

 

Pete

 

 

 

NVIDIA GeForce Titan X Pascal 12GB GDDR5X (900-1G611-2500-000)

by Nvidia

4.7 out of 5 stars   15 reviews from Amazon.com  |

Be the first to review this item

 

Price: £2,299.00 + £5.99 UK delivery

 

Note: Not eligible for Amazon Prime.

 

Only 2 left in stock.

Arrives before Christmas.

Get it as soon as 19 - 22 Dec. when you choose Standard Delivery at checkout. Details

Dispatched from and sold by Server and More GmbH & Co. KG.

 

2 new from £2,299.00

 

Not even free delivery for that price!

Well I had 24GB VRAM with my two (now "old") Titan X's

 

Hi Rob,

 

Could you clarify, please, the relevant amount of max VRAM available when 2 cards such as your "old" Titan X's, each with 12GB onboard VRAM, are working in SLI? My understanding is that the figure would be 12GB and not 24GB. Be nice if it was the latter but, somehow, I doubt it.

 

Regards,

Mike

 

 

Could you clarify, please, the relevant amount of max VRAM available when 2 cards such as your "old" Titan X's, each with 12GB onboard VRAM, are working in SLI?

 

I don't think "addressable" was used in any reference nor frame rendering?  As far as I know, only DX12 and compute applications can reference the full 24GB.  Sorry, no hints of future P3D.

 

But in my experience for the type of applications/games/sims I use, I've never needed more than 6GB VRAM even at 4K res.

 

Cheers, Rob.

Well Rob, you may not need believers but I always find your postings useful and informative so please keep them coming! :wink:    

 

As you say there are all kinds of variables at play that make it hard to identify real performance problem areas...and that's just with the settings in the sim itself (and NCP and NVI settings).  But even if two people match their sim settings, there are likely to be dozens of differences in their HW  and OS setups that will also cause differences in the performance results, so one has to be careful to draw the right conclusions.  But it's still good to have some common scenario setups to use as a starting point.

 

Kevin     

But in my experience for the type of applications/games/sims I use, I've never needed more than 6GB VRAM even at 4K res.

 

Cheers, Rob.

I've seen it recently and here's some other information.

 

http://www.tweaktown.com/tweakipedia/90/much-vram-need-1080p-1440p-4k-aa-enabled/index.html

 

The vast majority of cards out there have 4 GB VRAM or less.

 

Ran my standard benchmark test I've used over the years for P3D - see tests 26,34,35,36:

 

 

Here are the supporting test files:

 

ENAT Benchmark .cfg

https://drive.google.com/file/d/0Bw0Q-fAfEZwydHR0Sm0xZ3hoOE0/view?usp=sharing

 

ENAT Benchmark Scenario

https://drive.google.com/file/d/0Bw0Q-fAfEZwyQXRNN1htbWJGVk0/view?usp=sharing

 

FRAPS Logs

https://drive.google.com/file/d/0Bw0Q-fAfEZwyM2VJXy1GeFJzWGs/view?usp=sharing

 

However, two Titan X's in SLI did perform a little better when using SGSS AA vs. the single Titan X Pascal.

 

Cheers, Rob

 

 

Rob by any chance you have access to a VR Gear? Oculus or Vive, i was really curious to know how the Titan X Pascal performs in VR, dindt find anyone yet that as TXP in VR.